基于自适应EEMD和分层分形维数的风电机组行星齿轮箱故障检测

Diagnosis of Wind Turbine Planetary Gearbox Faults Based on Adaptive EEMD and Hierarchical Fractal Dimension

针对传统平均经验模态分解(EEMD)中添加白噪声参数需依据人工经验设定的缺陷,在研究引起模态混叠原因的基础上提出一种自适应EEMD方法。该方法可以根据信号本身特性,自适应设定白噪声标准差以达到最优分解效果。首先使用奇异值差分谱法对信号进行分解、重构,然后利用提取得到的高频冲击分量和噪声分量的复合分量对所需添加白噪声标准差大小进行自适应整定,最后通过自适应EEMD将信号分解为一系列本征模态函数(IMF)。分形维数对信号特征评价性能良好,所以用分形维数来识别不同类型振动信号是十分有效的。本文提出分层分形维数方法,可提高信号识别、分类效率和准确度。使用该复合方法处理仿真信号、风电机组传动系统实验平台信号均取得良好效果,证明了本文所提方法的有效性。

The main defect of the traditional ensemble empirical mode decomposition （EEMD） is that the important parameters of the added white noise are set by artificial experience. In the paper, an adaptive EEMD is proposed based on the study of the factors that caused the modal aliasing. This method could set the parameters for different signals adaptively to achieve optimal decomposition effectiveness. Firstly, the singular value decomposition （SVD） was used to decompose and reconstruct the signals. Next, the reconstruction signals were used to determine the parameters of the white noise adaptively. Finally, using the proposed method, the signals were decomposed to a series of intrinsic mode function （IMF）. Fractal dimension is good for the evaluation of the characteristics of IMF, so it is effective to identify different types of vibration signals. The hierarchical fractal dimension was used to improve the accuracy and efficiency of signal recognition. The experimental and simulation results of the gearbox of the wind turbine show that the proposed method is more effective compared with the existing techniques.